SkillTotal
← Back to home

State of AI Component Security

Open, reproducible data — not a black box. Run it yourself and get the same numbers.

Deterministic static scan of 32 real AI components (MCP servers, agent skills, and packages) — engine v0.17.0, ruleset 18, generated 2026-06-19. Every number is reproducible: the same public manifest through the same engine yields the same result. This characterizes the manifest; it is not a claim of statistical representativeness, and N grows as the manifest grows.

Risk level distribution

LOW
30 · 93.8%
MEDIUM
0 · 0%
HIGH
2 · 6.2%
CRITICAL
0 · 0%

0 of 32 (0%) carry a deliberate malicious indicator. Powerful capabilities are common; that is capability, not malice — they do not raise the score.

OWASP Agentic Skills Top 10

CategoryComponents%
AST0126.2%
AST02721.9%
AST03412.5%
AST0400%
AST0513.1%
AST0600%
AST0700%
AST0800%
AST0900%
AST1000%

AST06–AST10 are runtime/governance risks, not statically checkable, so they read 0 by construction. See the methodology mapping.

Capability prevalence

dynamic_code_execution26.2%
filesystem_read1340.6%
filesystem_write928.1%
install_time_execution721.9%
mcp_tools_detected1031.2%
network_egress1650%
shell_execution825%

Reproduce

Re-derive every number with the open-source engine and the public manifest:

pip install skilltotal
python tests/manual_eval/corpus_report.py

Full reportRaw JSONManifestMethodology